Lensfree Computational Microscopy Based On Multi-distance Phase Retrieval

Due to the variety of sample type and observation condition,high-throughput and portable microscope has been evolved into new demand for biomedical imaging.The traditional microscope using the refractive optical device is difficult to be assembled with a miniatured implementation.In contrast,lensfree microscopy dispenses of an objective lens and achieves an image reconstruction of the sample by post-processing of diffraction images,which shows a great promise to become a new generation of microscope that can accomplish a high-throughput and portable design with the help of intelligent computation.In this paper,we aim to design a series of algorithms,including image registration,convergence acceleration,data-efficiency enhancement,digital refocusing,pixel super resolution,color imaging,and single-frame imaging,to build two lensfree imaging platforms:one is used for whole-field and sub-pixel pathological imaging,and another for single-frame dynamic observation of living cell.The main content can be generalized as follows:An adaptive parameter correction method is proposed to solve the the parameter mismatch problem between real system and algorithmic model,which consists of tilt angle correction and diffractive distance estimation.For the correction of tilt illumination,the relation of tilt angle,lateral shift,and diffractive distance is derived in the near field region,and then a back-propagated image registration method based on cross-correlation peak shift is proposed to mitigate the lateral shift errors and align all defocused images.For the distance estimation,the influence of illumination mode,noise and tilt angle on foucisng curve is analyzed,and a new image sharpness quantification function based on the singular value fusion of the gradient is proposed to find the real sample-to-sensor distances.The corresponding experiment proves that this method is free of mechanical tuning and eliminates the image distoration induced by parameter mismatch.Dual-plane phase retrieval based on total variation constraint(DPR-TV)is proposed to improve the data-efficiency of lensfree pixel super-resolved system.In the pixel super-resolved system,a 3D-stack intensity images should be required for pixel super resolution and phase retrieval.Thus,the data-efficiency could be enhanced by decreasing the number of axial scanning plane.In the DPR-TV algorithm,an inverse problem is constructed by data fidelity and total variation norm.A gradient descent method based on weighted feedback and total variation minimization is desiged to solve the inverse problem.The corresponding experiment shows that the data amout and computational time are both decreased by DPR-TV algorithm,where the number of axial plane is decreased from 8 to 2.A single-frame phase retrieval based on nonlinear optimization(SFPR-NL)is proposed to achieve a high-fidelity image reconstruction for single-shot lensless imaging.In the SFPR-NL algorithm,a quadratic inverse problem of phase imaging is constructed by a physical prior of wave field propagation.A gradient-based method using fast iterative shrinkage thresholding optimization is designed to solve the nonlinear inverse problem.The corresponding experiments demonstrates that SFPR-NL algorithm could eliminate the twin-image artifact without the loss of imaging resolution,which increases the robustness of single-frame imaging.Based on the DPR-TV algorithm,a pixel super-resolved imaging platforms have been built,where a lateral resolution of 548nm and field-of-view of 28.6mm~2 are achieved.In comparison with conventional method(8×6×6),the data amout is decreased from 288 to 50 by DPR-TV algorithm.Based on the SFPR-NL algorithm,a portable platform is established by a 3D printer.For the static target,the experiments of the pathological slides and label-free microglia cell demonstrate that the portable platform acomplishes the high-quality reconstruction of dense sample.For the dynamic target,our platform could accomplish incubator-based imaging of living He La cell,where the cell migration and death is observed.